| 02:18 | Lachy | is experimenting with IRC BNC servers... |
| 02:19 | <Lachy> | I'm using psybnc right now, but it's overly complicated and not working well. Does anyone know of a better alternative? |
| 03:26 | Hixie | ran into an rdfa person earlier |
| 03:27 | <Hixie> | i'm now more convinced than before that html5 is following the right course in terms of rdfa, distributed extensibility, etc |
| 03:29 | <Lachy> | Hixie, what did the rdfa person tell you? |
| 03:30 | <Hixie> | that html5 was a horrible mistake, basically |
| 03:30 | <Hixie> | we had a long 40 minute discussion |
| 03:30 | <Lachy> | can you say who it was? |
| 03:31 | <jcranmer> | 40 minutes to say "your project SUCKS" ? |
| 03:31 | <Hixie> | Hal Abelson |
| 03:32 | <Lachy> | ok, I don't know him. |
| 03:32 | <Hixie> | jcranmer: no, we were discussing things like whether there was any reason to believe that authors were interested in micro-level detailing of data types (as in microformats or rdfa), etc |
| 04:20 | <BenMillard> | Hixie, are you here still? |
| 04:22 | <BenMillard> | oh, you're AFK |
| 04:48 | <jacobolus> | Hixie: the Hal Abelson who taught SICP with Sussman for all those years? |
| 04:55 | <jacobolus> | funny that hearing objections would make you more convinced. :) I guess the objections were not compelling? :) |
| 05:44 | <Hixie> | jacobolus: yeah, that Hal. It wasn't so much that the objections made me more convinced, so much as articulating the counterarguments made me realise how strong the arguments in favour of our current course were compared to the arguments against |
| 05:54 | <BenMillard> | Hixie, at a very rough count it looks like I've bookmarked (very approximately) 900 web pages which do something interesting with their markup |
| 05:54 | <BenMillard> | each bookmark has maybe a sentence describing why I bookmarked it |
| 05:55 | <BenMillard> | I indent to scrape those entries out Firefox's bookmarks.html file and then filter the markup into something halfway sane for the web |
| 05:55 | <BenMillard> | this should hopefully meet the midpoint criteria for my Mozilla sponsorship on 1st September |
| 05:56 | <BenMillard> | after that, I can tailor the document to make it more useful for HTML5 work |
| 05:59 | <Hixie> | cool |
| 06:23 | <hdh> | "The insertRow(index) method must, when invoked on an element table section, act as follows:" should read "table section element"? |
| 06:24 | <Hixie> | no, "table section" is a variable in that sentence |
| 06:25 | <Hixie> | "The insertRow(index) method must, when invoked on an element /table section/, act as follows" |
| 06:25 | <hdh> | ah ok |
| 07:05 | <hsivonen> | zcorpan: nope. not intended. I wrote the script kick-off code in the browser DOM tree builder for HTML script and forgot that SVG script can have non-text children |
| 07:05 | <hsivonen> | zcorpan: thanks |
| 08:02 | <hsivonen> | zcorpan: http://livedom.validator.nu/?%3C!DOCTYPE%20html%3E%0D%0A%3Csvg%3E%3Cscript%3Ea%3Cfoo%3Eb%3C%2Fscript%3E fixed. thanks. |
| 08:54 | <zcorpan> | hsivonen: oh sorry. the <script>a<foo>b thing seems to be a bug in opera |
| 08:56 | <hsivonen> | zcorpan: no, it was a bug in the parser |
| 08:56 | <hsivonen> | 10:05 < hsivonen> zcorpan: |
| 08:56 | <hsivonen> | http://livedom.validator.nu/?%3C!DOCTYPE%20html%3E%0D%0A%3Csvg%3E%3Cscript%3Ea%3Cfoo%3Eb%3C%2Fscript%3E fixed. thanks. |
| 08:56 | <hsivonen> | or rather, it was a bug in the script kick-off code in the GWT tree builder |
| 08:57 | <Hixie> | so i was talking to the translation guys at google earlier, regarding the no-translate thing |
| 08:57 | <Hixie> | and i was amused to find out that we just ignor lang="" altogether for the purposes of translation |
| 08:57 | <hsivonen> | hmm. now that I think about it, there's a pretty serious bug in the case of <script><script> in SVG |
| 08:57 | <Hixie> | because it's more accurate to just autodetect the language than to use the metadata |
| 08:58 | <hsivonen> | yay for metadata |
| 08:58 | <hsivonen> | Hixie: how should scripts be run when SVG scripts elements nest? |
| 08:58 | <Hixie> | ask the svgwg |
| 09:00 | <hsivonen> | their recent tendency seems to be towards doing what HTML browsers do with scripts |
| 09:04 | <Hixie> | sounds good to me |
| 09:04 | <Hixie> | so do the same as what XHTML <script>s do when nested |
| 09:05 | <hsivonen> | what happens with those? |
| 09:06 | <hsivonen> | (I should probably know, but I don't.) |
| 09:06 | <Hixie> | search for "When a script block is inserted into a document, the user agent must act as follows" |
| 09:06 | <Hixie> | though i guess the spec as written makes xml script elements not work |
| 09:06 | <heycam> | hsivonen, let me know what happens for that so i can append it to my SVG <script> fixes |
| 09:06 | <heycam> | (if you could) |
| 09:07 | <hsivonen> | heycam: ok. |
| 09:07 | <Hixie> | since the xml parser (where is it defined? oh well) presumably inserts the script element before the contents are inserted |
| 09:07 | <Hixie> | and thus the scripts never actually execute |
| 09:07 | <Hixie> | i guess that's a bug that should be fixed |
| 09:07 | <Hixie> | not sure how to fix it though |
| 09:07 | <Hixie> | maybe script elements shouldn't be inserted until they're all parsed |
| 09:08 | <Hixie> | or there should be something that happens when the end tag is parsed |
| 09:08 | <Hixie> | the latter seems best |
| 09:08 | <Hixie> | file a bug |
| 09:08 | <hsivonen> | ok |
| 09:14 | <zcorpan> | hsivonen: ah. i was confused by opera's agressive caching again |
| 10:24 | <Lachy> | I upgraded several whatwg blog accounts to Editor, giving them the ability to moderate posts, manage categories, etc. |
| 10:52 | <Lachy> | Hixie, yt? |
| 10:53 | <Lachy> | Hixie, can you send me a copy of the latest acid3 test. It appears my copy is out of date now. |
| 10:56 | <Lachy> | actually, could you set up a script that creates a tarball of all the latest files when I go to acid3.acidtests.org/latest.tar.gz (or whatever) |
| 12:27 | <zcorpan> | hsivonen: <svg></body> makes livedom hang |
| 12:29 | <hsivonen> | zcorpan: that's weird. thanks. |
| 12:29 | <hsivonen> | http://parsetree.validator.nu/?parser=html5&content=<svg><%2Fbody>; works |
| 12:30 | <zcorpan> | indeed |
| 12:33 | <zcorpan> | hsivonen: is it intentional to drop comments in livedom? |
| 13:12 | <hsivonen> | hmm. <svg></body> doesn't hang in Hosted Mode |
| 13:26 | <hsivonen> | zcorpan: I'm suspecting an unfixed variant of the same compiler bug that was more obvious earlier |
| 13:26 | <hsivonen> | but this is hard to debug, because now there's a new compiler bug that makes debugging hard |
| 14:05 | <Lachy> | JohnResig, does your selectors api test suite test elements outside of the document? |
| 14:06 | <JohnResig> | Lachy: yes |
| 20:26 | <Philip`> | http://wiki.whatwg.org/wiki/User_talk:Highhi - spam |
| 20:52 | <Philip`> | takkaria: The image on http://takkaria.org/dmoz/ seems purely decorative and therefore must have alt="" |
| 21:05 | <Philip`> | "it's more accurate to just autodetect the language than to use the metadata" - wouldn't it be even more accurate to use the metadata as an input to the autodetection heuristics, since there's bound to be a non-zero correlation between the metadata and the data and therefore it's better to use that input than to ignore it? |
| 21:22 | <Dashiva> | How certain is that? |
| 21:25 | <Philip`> | How certain is what? |
| 21:25 | <Dashiva> | That the correlation is positive |
| 21:28 | <Philip`> | Pretty certain - if lang is currently entirely random and unrelated to the actual language, then I just need to create a single web page which uses lang correctly, and then there'll be a positive correlation, and the only way it'd get back to zero correlation is if someone maliciously uploads a new page with an intentionally incorrect lang, which seems pretty unlikely |
| 21:29 | <Philip`> | (and if they upload two incorrect pages, then there'll be a negative correlation, which you could usefully feed into your language autodetection heuristics so that your choice is biased away from the metadata because it's most likely to be incorrect) |
| 21:29 | <Dashiva> | But we'd have to know which |
| 21:30 | <Dashiva> | There could be page creators with English hardcoded, e.g. |
| 21:31 | <Philip`> | If you have a million pages which say lang="en" because it's hardcoded into the editor and is unrelated to the actual language, and one page that says lang="en" because it's actually English, then the probability of a page being English given that it has lang="en" is greater than the probability of it being English given no prior knowledge |
| 21:32 | <Philip`> | kangax: If I'm remembering correctly that you asked about Rhino canvases, you might want to see http://rhino-canvas.sourceforge.net/ |
| 21:33 | <kangax> | Philip`: yes, I've seen this, but when running couldn't make it produce any result |
| 21:33 | <Dashiva> | It'd have to be significant enough to be useful |
| 21:33 | <Philip`> | Dashiva: (...assuming the distribution of languages on the pages created with the lang="en" editor is identical to the distribution of languages on pages without lang, or something like that) |
| 21:34 | <Philip`> | Dashiva: That depends solely on your definition of "useful" |
| 21:34 | <Dashiva> | Useful in that there exist pages that are ambiguous without it and not with it |
| 21:34 | <kangax> | Philip`: eventually I found out that Jaxer (running on Rhino) works with canvas nicely |
| 21:34 | <takkaria> | Philip`: fixed, ta |
| 21:34 | <kangax> | Philip`: and implements toDataURL which I needed |
| 21:36 | <Philip`> | Dashiva: If providing extra information results in your autodetection algorithm having a one-in-a-million chance of giving a better result, then that seems useful if you're looking at more than a million pages, and on the web you probably are :-) |
| 21:36 | <Dashiva> | But it probably isn't that much |
| 21:36 | <Dashiva> | It has to be a million pages with ambiguous content |
| 21:37 | <Philip`> | kangax: Ooh, Jaxer has a canvas implementation? |
| 21:37 | <Dashiva> | And even then there's the cost of figuring out how much weight to give it, and implement the weighting, and suffer from false positives when you overweight :) |
| 21:37 | <kangax> | Philip`: yep |
| 21:37 | <kangax> | Philip`: it managed to run all of my code (which is getting quite complex) |
| 21:38 | <Philip`> | kangax: If I wasn't too lazy (which I am), I'd update http://philip.html5.org/tests/canvas/suite/tests/results.html to include that implementation, which could be interesting :-) |
| 21:39 | <kangax> | they are using gecko if i'm not mistaken |
| 21:39 | <kangax> | not sure which, though |
| 21:39 | <kangax> | (not sure which version) |
| 21:40 | <Philip`> | Oh... If it's "running on Rhino", I presume that means it's doing stuff in Java, so how does it make use of Gecko? |
| 21:42 | <kangax> | I think via XUL |
| 21:43 | <kangax> | Philip`: http://www.aptana.com/node/266 |
| 21:43 | <kangax> | "Jaxer is built on the same technology as Firefox which means we can take advantage of XUL-like functionality on the server." <-- I assume they are talking about Gecko |
| 21:47 | Philip` | fails to find an obvious description of technical details on what it does |
| 22:03 | <Philip`> | Hixie: Are you aware that the commit-watchers emails are saying "/home/ianh/svn/webapps/hooks/commit-email.pl: `/usr/bin/svnlook diff /home/ianh/svn/webapps -r 2073' failed with this output:"? |
| 22:12 | <Hixie> | Philip`: i was not, investigating, thanks |
| 22:12 | <Hixie> | Philip`: re the language thing, my understanding is that the heuristics are so significantly more accurate than lang="" that looking at lang="" wouldn't help any measurable amount |
| 22:15 | <Hixie> | so er |
| 22:15 | <Hixie> | the problem is svn is segfaulting |
| 22:15 | <Hixie> | wtf |
| 22:20 | <Hixie> | well i have no idea how to fix that |
| 22:20 | <Hixie> | so i guess we'll have to live with it for now |
| 22:32 | <gsnedders> | Hixie: can you reply to my email about datagrid ASAP? |
| 22:32 | <Hixie> | sent today? |
| 22:32 | <gsnedders> | Hixie: Just now. |
| 22:32 | <gsnedders> | I only just got internet back from being on holiday :) |
| 22:33 | <Hixie> | The interactive content vs element thing? |
| 22:33 | <gsnedders> | yeah |
| 22:33 | <gsnedders> | (I need it so I can release spec-gen 1.0RC1 tomorrow, or maybe Sun) |
| 22:33 | <Hixie> | the answer to the question is yes |
| 22:33 | <Hixie> | but editing the spec will be a bitch cos i'm in the middle of a bazillion edits |
| 22:34 | <gsnedders> | Hixie: Thanks. That'll do for me. :P |
| 22:34 | <Hixie> | do you need the spec edited? |
| 22:34 | <gsnedders> | Nope. |
| 22:34 | <Hixie> | ok cool |
| 22:34 | <Hixie> | yeah that was just a typo |
| 22:34 | <gsnedders> | (It obviously ought to be fixed, but I don't have any need for that to be done quickly) |
| 22:34 | <Hixie> | i always think in terms of "bla element" like flow element, phrasing element, etc |
| 22:34 | <gsnedders> | Yeah, I was guessing that. |
| 22:34 | <Hixie> | but i have to call it content to allow text nodes to fall into it, etc |
| 22:34 | <Hixie> | so i keep making that mistake |
| 22:35 | <Hixie> | anything else i can help you with? |
| 22:35 | <gsnedders> | I think not with |
| 22:35 | <gsnedders> | Have we published the Aug draft MikeSmith was talking of? |
| 22:35 | <takkaria> | not yet |
| 22:36 | <gsnedders> | All right. |
| 22:36 | <Hixie> | i'll be switching to spec gen asap after we do, if it's available :-) |
| 22:37 | <gsnedders> | spec-gen 1.0b2 is perfectly bug-free, spec-gen 1.0RC1 will cause ID gen breakages, though |
| 22:37 | <gsnedders> | and RC1 really is an RC :) |
| 22:37 | <Hixie> | hehe |
| 22:37 | <Hixie> | you sure are taking this seriously :-P |
| 22:38 | <gsnedders> | :P |
| 22:38 | <gsnedders> | Hixie: Or maybe I'm just a perfectionist :P |
| 22:39 | gsnedders | ponders how to get around the issue that he uses XPath's normalize-space() which follows XML's S, and not HTML 5 whitespace |
| 22:40 | <Philip`> | I suggest getting around the issue by not caring about it |
| 22:40 | <Hixie> | "a documented limitation of this software..." |
| 22:41 | <gsnedders> | All that that affects is the whitespace in the middle of "latest version" for parsing the W3C status :P |
| 22:41 | <gsnedders> | s/af/ef/ |
| 22:46 | <gsnedders> | I know I could fix it, it's just a question of with a how big of a perf. hit |
| 22:54 | <gsnedders> | (The ID gen changes only really effect things that play around with edge cases, like HTML5 :D) |
| 22:54 | <Hixie> | heh |
| 23:15 | <Hixie> | i guess we'll have to study how people use Content-Language |
| 23:31 | <gsnedders> | Hixie: As an HTTP header? |
| 23:31 | <Hixie> | meta |
| 23:32 | <gsnedders> | I can give you data for HTTP header, but not meta :P |